Good morning! Today we're covering Apple's official Gemini partnership that sent Alphabet past $4 trillion, Anthropic's new no-code AI agent for everyday work, Meta's Reality Labs pivot, regulators worldwide cracking down on deepfake tools, and ElevenLabs' explosive growth numbers. Let's dive in.

After months of speculation, Apple and Google confirmed yesterday that Gemini will power certain Apple Intelligence features, including enhanced Siri capabilities. The announcement sent Alphabet's market valuation soaring past $4 trillion for the first time, making it only the second company after Apple to reach this milestone.

The partnership gives Apple access to Google's most advanced Gemini models to handle complex queries beyond what Apple's on-device AI can manage. Users will be able to opt into Gemini-powered features through their Apple Intelligence settings, with the integration expected to roll out gradually through iOS updates. This marks a significant strategic shift for Apple, which has historically avoided such deep partnerships with competitors.

The deal represents both companies hedging their bets in the AI race. Apple gains immediate access to cutting-edge language models without the massive infrastructure investment required to train them, while Google secures distribution across Apple's billion-plus devices. For users, it means Siri should finally be able to handle more sophisticated conversational tasks and complex questions. The financial terms weren't disclosed, but analysts expect Google will pay Apple similarly to their existing search agreement.

Read the full story →

Anthropic released Cowork yesterday, a new feature that transforms Claude into a desktop AI agent that can read, edit, and organize files on your computer—no coding required. Unlike previous AI coding tools that required technical expertise, Cowork is designed for everyday knowledge workers who need help managing documents, spreadsheets, and other files.

The feature works through Claude Desktop and gives the AI controlled access to specific folders you designate. You can ask Claude to find information across multiple documents, reorganize files by project, extract data from spreadsheets, or even draft summaries by pulling information from various sources. Anthropic emphasized that users maintain complete control over which folders Claude can access, and the AI can't take actions without explicit permission for each task.

This represents a significant step toward practical AI agents that augment daily work rather than replace specialized tasks. The timing is notable—Anthropic simultaneously announced C-suite changes to expand its internal incubator, bringing on Instagram co-founder Mike Krieger to lead new product development. The company is clearly betting that the next wave of AI value will come from agents that integrate seamlessly into existing workflows rather than requiring users to adapt to new interfaces.

Read the full story →

Meta confirmed layoffs in its Reality Labs division yesterday as the company pivots resources toward augmented reality wearables and away from some VR projects. While Meta didn't specify exact numbers, the restructuring represents a strategic shift after Reality Labs has burned through over $50 billion since 2020 with limited commercial success.

The company is redirecting investment toward AR glasses and smart wearables, following the success of its Ray-Ban Meta smart glasses which have gained traction where previous products stumbled. This marks a notable retreat from Mark Zuckerberg's metaverse ambitions in favor of more practical AI-powered devices that integrate into daily life. The shift aligns with broader industry trends showing consumers prefer lightweight AR experiences over immersive VR headsets for regular use.

For the thousands of Reality Labs employees, this represents yet another wave of uncertainty at Meta, which has conducted multiple rounds of layoffs since 2022. The company said affected employees will receive severance and job placement assistance. Meanwhile, Meta continues massive investments in AI infrastructure, announcing its own data center initiative to support AI training—suggesting the company views AI software as more immediately valuable than hardware platforms.

Read the full story →

Multiple governments escalated actions against X's Grok AI chatbot yesterday, with the UK's communications regulator Ofcom launching a formal investigation and Malaysia blocking access entirely. The controversy centers on Grok's ability to generate realistic explicit images of real people without consent—a capability most other AI companies have restricted.

UK officials specifically cited the platform's failure to prevent the creation of sexualized AI-generated images, particularly after widespread reports of abuse targeting women and minors. Ofcom's investigation will examine whether X violated the Online Safety Act, which requires platforms to prevent illegal content and protect users from harm. Malaysia's communications minister took more immediate action, ordering internet service providers to block Grok access after the tool was used to create fake explicit images, calling it a threat to digital safety.

The crackdown highlights growing global frustration with inconsistent AI safety standards. While companies like OpenAI, Anthropic, and Google have implemented strict content filters, Elon Musk has positioned Grok as maximally permissive in the name of free expression. The U.S. Senate also passed the DEFIANCE Act yesterday, giving victims of nonconsensual deepfakes the right to sue creators and platforms. Experts warn this may be just the beginning of AI-generated abuse targeting women.

Read the full story →

ElevenLabs CEO Mati Staniszewski revealed yesterday that the voice AI startup crossed $330 million in annual recurring revenue in 2025—a staggering milestone for a company that was valued at $3 billion just months ago. The announcement underscores explosive demand for AI-generated voice technology across industries from entertainment to customer service.

ElevenLabs has become the go-to platform for realistic voice synthesis, offering tools that can clone voices, generate speech in dozens of languages, and create entirely new synthetic voices. The company's technology powers everything from audiobook narration to video game characters to multilingual content localization. Major publishers, studios, and content creators have adopted the platform as it offers dramatically faster and cheaper production compared to traditional voice recording.

The rapid revenue growth—from zero to $330M ARR in roughly two years—demonstrates how quickly AI tools can scale when they solve real production bottlenecks. It also reflects the broader shift toward AI-native workflows in creative industries. ElevenLabs has been careful to implement consent and authentication features after early concerns about voice cloning abuse, requiring verification for voice cloning and offering detection tools. The company is reportedly considering raising additional funding at a significantly higher valuation.

Read the full story →

Google AI quietly released MedGemma 1.5 this week, the latest update to its open medical AI models designed specifically for healthcare developers. The updated models offer improved accuracy on medical question answering, clinical note summarization, and health information extraction tasks—all while maintaining the open-source licensing that made the original MedGemma popular with researchers.

MedGemma 1.5 builds on Google's Gemini foundation models but has been extensively fine-tuned on medical literature, clinical guidelines, and healthcare datasets. The models come in multiple sizes to support different deployment scenarios—from lightweight versions that can run on edge devices to more capable versions for cloud-based applications. Google emphasized that these are research models not intended for direct clinical use without proper validation and regulatory approval.

The release comes as both Anthropic and OpenAI announced specialized healthcare AI offerings this week, signaling that medical applications represent a major battleground for AI companies. Healthcare offers enormous potential value from AI—better diagnosis support, reduced administrative burden, personalized treatment—but also carries unique risks and regulatory requirements. By providing open models, Google is betting that enabling a broad ecosystem of developers will accelerate innovation while building goodwill in the medical community.

Read the full story →

With Apple now officially bringing Google's Gemini into Siri, and Anthropic releasing Cowork to manage your local files, AI agents are getting seriously close to our daily workflows and personal data. I'm curious: what level of access would you actually feel comfortable giving an AI assistant? Would you let Claude reorganize your documents? Would you trust Gemini-powered Siri with sensitive queries? Where do you personally draw the line? Hit reply and let me know—I genuinely read every response and often feature interesting perspectives in future newsletters!

That's all for today! If you found this valuable, forward it to a colleague who needs to stay current on AI. And if someone forwarded this to you, subscribe at dailyinference.com for daily AI news in your inbox.

— The Daily Inference Team